An integrated palliative and respiratory care service for patients with advanced disease and refractory breathlessness: a randomised controlled trial

Summary Background Breathlessness is a common and distressing symptom, which increases in many diseases as they progress and is diﬃ cult to manage. We assessed the eﬀ ectiveness of early palliative care integrated with respiratory services for patients with advanced disease and refractory breathlessness. Methods In this single-blind randomised trial, we enrolled consecutive adults with refractory breathlessness and advanced disease from three large teaching hospitals and via general practitioners in South London. We randomly allocated (1:1) patients to receive either a breathlessness support service or usual care. Randomisation was computer generated centrally by the independent Clinical Trials Unit in a 1:1 ratio, by minimisation to balance four potential confounders: cancer versus non-cancer, breathlessness severity, presence of an informal caregiver, and ethnicity. The breathlessness support service was a short-term, single point of access service integrating palliative care, respiratory medicine, physiotherapy, and occupational therapy. Research interviewers were masked as to which patients were in the treatment group. Our primary outcome was patient-reported breathlessness mastery, a quality of life domain in the Chronic Respiratory Disease Questionnaire, at 6 weeks. All analyses were by intention to treat. Survival was a safety endpoint. This trial is registered with ClinicalTrials.gov, number NCT01165034. Findings Between Oct 22, 2010 and Sept 28, 2012, 105 consenting patients were randomly assigned (53 to breathlessness support service and 52 to usual care). 83 of 105 (78%) patients completed the assessment at week 6. Mastery in the breathlessness support service group improved compared with the control (mean diﬀ erence 0∙58, 95% CI 0∙01–1∙15, p=0∙048; eﬀ ect size 0∙44). Sensitivity analysis found similar results. Survival rate from randomisation to 6 months was better in the breathlessness support service group than in the control group (50 of 53 [94%] vs 39 of 52 [75%]) and in overall survival (generalised Wilcoxon 3·90, p=0∙048). Survival diﬀ erences were signiﬁ cant for patients with chronic obstructive pulmonary disease and interstitial lung disease but not cancer. Interpretation The breathlessness support service improved breathlessness mastery. Our ﬁ ndings provide robust evidence to support the early integration of palliative care for patients with diseases other than cancer and breathlessness as well as those with cancer. The improvement in survival requires further investigation.


Introduction
Breathlessness is a common and distressing symptom in many advanced chronic diseases, causing considerable disability, anxiety, and social isolation. [1][2][3] Worldwide, more than 75 million people have breathlessness every year, including more than 90% of the 65 million people with severe lung disease, 4 more than 50% of the 10 million with incurable cancer, and 50% of the 23 million with heart failure. 5,6 Breathlessness increases as the disease progresses, 7 is frightening for patients and families, and often results in emergency hospital admission because it is accompanied by feelings of loss of mastery over breathing and panic. 7,8 Once treatment of the underlying disease is optimised, breathlessness that continues is deemed refractory. 1 Patients with refractory breathlessness in advanced disease have many symptoms and concerns that are complex and interact; consequently palliative care has been recommended. 4,9 In this study, we developed and assessed a new short-term breathlessness support service. This provided one point of access for patients, brought together palliative care and respiratory medicine, and responded to the call for shared care at an earlier stage in disease than usual. 10,11 We hypothesised that patients attending the breathlessness support service, compared with those receiving standard care, would have better mastery of breathlessness at 6 weeks.

Study design and participants
This trial was a randomised controlled, parallel group, pragmatic, single-blind fast-track trial in South London, UK, recruiting patients between Oct 22, 2010 andSept 28, 2012. We screened for potential patients across three large teaching hospitals and via general practitioners.
Patients were included according to a standard proforma completed by the identifying clinician. Patients had to meet all criteria: refractory breathlessness on exertion or rest (MRC dyspnoea scale score ≥2), despite optimum treatment of the underlying disease, as deemed by the identifying clinician; advanced disease such as cancer, chronic obstructive pulmonary disease (COPD), chronic heart failure, interstitial lung disease, and motor neuron disease; willing to engage with shortterm home physiotherapy and occupational therapy; and able to provide informed consent. Patients were excluded for any of the following: breathlessness of unknown cause; a primary diagnosis of chronic hyperventilation syndrome; completely house (or hospital or nursing home) bound, despite off er of free transport to clinic; or within 2 weeks of treatment for an acute exacerbation. Such patients were reapproached after 2 weeks.
Protocol, procedures, information sheets, consent forms, and questionnaires were approved through the inde pendent UK Integrated Research Approval System via the ethics committee at King's College Hospital (Ref. 10/H0808/17). We then applied for and were granted NHS Research and Development approval in all recruiting sites. Patients gave written informed consent before enrolment. Our protocol 12 followed CONSORT recommendations. There were no protocol deviations.

Randomisation and masking
Using data from the baseline interview, the King's Clinical Trials Unit's Online randomisation system, independent of research and clinical teams, randomly assigned (1:1) patients to the intervention (immediate access to breathlessness support service in addition to standard care) or control group (standard best practice; off ered breathlessness support service after 6 weeks). Allocation was done by minimisation 13 to balance four potential confounders: cancer versus non-cancer, breathlessness severity (numerical rating scale >3 or not), presence (or not) of an informal caregiver, and ethnic origin (white or other). After randomisation, the clinical trials unit team informed the breathlessness support service clinic administrator of the patient's study group via secure email, who then arranged clinic appointments accordingly. Research nurses and interviewers were masked to treatment allocation. Patients were aware of treatment allocation, and were asked not to disclose this information to interviewers or research nurses. The trial coordinator and the trial administrator were aware of treatment allocation; the coordinator informed the research nurses when, and with whom, they had to do interviews.

Procedures
The breathlessness support service is an additional service to usual UK National Health Service (NHS) care. It is a multi-professional integrated service that combines respiratory, physiotherapy, occupational therapy, and palliative care assessment and management. It brings together assessment and treatment of physical, emotional, psychological, and spiritual concerns, through one point of access. The service comprises (appendix pp 1-2) a fi rst outpatient clinic appointment with respiratory medicine and palliative care clinicians assessing present treatment and concerns. The patient (and family if present) is given a breathlessness pack including information, management, and pacing guidance, a hand-held fan or water spray, and a poem (a short mantra to help breathing and relaxation during crises) and helped to agree a crisis plan. A home assessment is done 2-3 weeks after the clinic visit by a physiotherapist and/or occupational therapist to assess the need for walking and home aids and adaptations, reinforcement of self-management, and further guidance on pacing and exercises, including a DVD when appropriate. 4 weeks after the fi rst clinic visit, a second and fi nal clinic appointment with a palliative care specialist is arranged to agree further actions and a discharge plan.
Service modelling for the breathlessness support service is built on the nurse-led clinic, developed by Bredin and colleagues 14 and the palliative care and physiotherapy approach developed by Booth and colleagues, 15 and systematic reviews, 16 qualitative interviews, 2 crosssectional 9,17 and longitudinal studies, 18 and consultation with local stakeholders. 12 These data suggested that breathlessness support services should provide one point of access, integrate palliative care with existing services,  19,20 • Severity of breathlessness in the previous 24 h on a 0-10 numerical rating scale (NRS), average, at rest, and on exertion 12 • London Chest Activity of Daily Living, a questionnaire of the level of disability induced by breathlessness for 15 activities (in four areas: personal care, domestic, physical, and social); each activity is scored 0-5 (0=I wouldn't do it anyway, 5=someone else needs to carry out the activity) 12 • EQ-5D and EQ-VAS which assess mobility, self-care, usual activities, pain or discomfort, anxiety or depression according to three levels of severity (1=no problems, 2=some or moderate problems, and 3=extreme problems), plus a Visual Analogue Scale (VAS) of present health-related quality of life, scored 0-100 12 • Palliative care Outcome Scale, a ten-item measure for advanced disease widely validated in cancer and non-cancer; each item is rated 0 (no problem) to 4 (overwhelming problem) 21 • Hospital Anxiety and Depression Scale (HADS), a 14-item measure of psychological distress with separate anxiety and depression subscales 12 • Client Services Receipt Inventory (CSRI) 12 in which patients reported the health, voluntary, and social care services received over the past 3 months, or if follow-up since the last research interview 12 off er outpatient and home contact, and focus on improving patient self-management. Patients randomly assigned to the control group continued with optimum management as provided by their usual services in accordance with relevant UK guidance to ensure best practice (appendix pp [3][4][5]. After the 6 week (primary endpoint) research interview, these patients were off ered the breathlessness support service.
Study measurements included the Chronic Respiratory Disease Questionnaire, severity of breathlessness in the previous 24 h, the London Chest Activity of Daily Living questionnaire, EQ-5D and EQ-VAS, the Palliative care Outcome Scale, the Hospital Anxiety and Depression Scale (HADS), and the Client Services Receipt Inventory (panel 1). These measurements were collected in a standard questionnaire booklet consisting of demographic, clinical, outcome assessments, and use of health-care services. Research data were collected in face to face interviews with patients, usually in their own homes, at baseline and 6 weeks follow-up (the primary endpoint). In addition face to face qualitative interviews were conducted after the trial was completed.
At baseline and 6 weeks follow-up, interviewers measured pulmonary function and oxygen saturation with a portable spirometer and fi nger pulse oximeter.

Outcomes
The primary outcome was breathlessness mastery at 6 weeks as recorded in the 6 week face to face interview, determined according to one domain of the quality of life measure, the Chronic Respiratory Disease Questionnaire 12,19 (panel 1). Mastery is the average of four questions about the feeling of control over the disease and its eff ects on quality of life and function (range=1 [maximum impairment] to 7 [no impairment]). Secondary outcomes included: severity of breathlessness on exertion in the previous 24 h, activity (assessed by London Chest Activity of Daily Living questionnaire), other domains of the Chronic Respiratory Disease Questionnaire (breathlessness, fatigue, and emotional function), quality of life (EQ-5D), palliative needs (assessed by Palliative care Outcome Scale), depression and anxiety (measured by the Hospital Anxiety and Depression Scale [HADS]), and spirometry. Patient survival (since randomisation) was planned to be calculated from dropout and missing data, with exact dates extracted from clinical records, as a safety secondary endpoint, as recommended in clinical trials. 22 We also planned to collect and analyse caregiver burden data. But it proved diffi cult to contact lay caregivers for consent because many used the research nurse visit as an opportunity to go out while someone was with the patient. Only 11 caregivers were interviewed at baseline and only four at 6 week follow-up. Therefore these data were not analysed further. 20 patients were purposefully selected for qualitative interviews, after completing the trial, to include a mix of ages, sex, diagnoses, and presence of caregiver. The interviews were semi-structured and followed a topic guide on patients' expectations, experiences, and views about the content, format, and eff ect of the breathlessness support service. Questions were open-ended and not based on pre-existing theory. Interviews were taperecorded and transcribed verbatim.

Statistical analysis
On the basis of our primary outcome, the Chronic Respiratory Disease Questionnaire mastery domain, we estimated that more than 34 patients per group would detect a mean diff erence of 0•70 (SD 1), a p value of less than 0•05 at power 80%. 19,20 To allow for a conservative estimated attrition of 40% we planned to recruit at least 110 patients into the study.
All randomly assigned participants were included in the intention-to-treat analysis. Missing data were explored according to cause. 23 Figure 1: Trial profi le *Our biggest loss before consent was the 47 individuals (21·8% of 216 referred to the trial, and 42·3% of those not consented and randomly assigned) whom we were unable to contact. Appendix p 9 shows the eff orts made to contact people.
prespecifi ed, we used independent samples Student's t test to compare patient mastery (primary outcome) and secondary outcomes at 6 weeks, by trial group. Sensitivity analysis explored the robustness of results: fi rst, accounting for diff erences in patient diagnoses and baseline scores using analysis of covariance (ANCOVA); second, with multiple imputations of missing data; 23 and third with pre-post analysis of breathlessness support service and control groups. Survival was calculated from date of randomisation to date of death and plotted using the Kaplan-Meier method. Patients who were still alive on Sept 1, 2013, after the last patient was recruited were censored at this date. We calculated the survival rates to 180 days (6 months) from consent in both groups. With all data to 1 year from last recruitment, we assessed overall survival diff erence between two groups with the generalised Wilcoxon (Breslow) test, which is more sensitive in the detection of early diff erences in survival, 24 which is important in the fast-track or wait-list design.
We calculated costs by combining Client Service Receipt Inventory data with UK 2011-12 unit costs. 25 Cost data are usually skewed; therefore, we used a bootstrapped regression model to produce confi dence intervals. Statistical signifi cance was accepted for p values less than 0•05.
Qualitative interviews were imported into NVivo version 7 and content analysis explored patients' own views and experiences of the breathlessness support service, and in particular how issues related to the primary and secondary outcomes. We created categories inductively, with attention to terms and content, from the interview data. Through a process of constant integration of categories and their properties, or constant comparison, the fi ndings became relevant at a more abstract level. We used simple counting when possible to discover more defi nite patterns in views. Member checks and teamwork were used to establish credibility.
This trial is registered with ClinicalTrials.gov, number NCT01165034.

Role of the funding source
The funders of the study had no role in study design, data collection, data analysis, data interpretation, or writing of the report. The corresponding author had full access to all the data in the study and had fi nal responsibility for the decision to submit for publication.    for the HADS, the mean scores were 9 for anxiety and 10 for depression, both above the cutoff for clinical signifi cance. At week 6, 82 of 105 (78%) patients completed assessments. The main reasons for attrition were illness or death (fi gure 1). Attrition to the primary outcome was slightly lower than estimated (22% not 40%); therefore, we agreed to stop recruitment after 105 patients had consented. Missing data, death, and dropout were not associated with baseline mastery score, FEV 1 or other key variables except oxygen saturation (appendix p 7). Four patients died by week 6, two had cancer (one in breathlessness support service group), and two had interstitial lung disease (both in control group).

Between
We recorded a signifi cant improvement in the primary outcome, the mastery domain of the Chronic Respiratory Disease Questionnaire, in the breathlessness support service group compared with the control group at 6 weeks (table 2). Patients receiving the breathlessness support service had on average a 16% improvement for breathlessness mastery over the control group (mean diff erence 0·58, eff ect size 0·44, control group mean score 3·57). Results were similar to those from our sensitivity analysis of the primary outcome: ANCOVA adjusted for diagnosis, p=0•037; ANCOVA adjusted for diagnosis and baseline score, p=0•05; multiple imputation (number of imputations 45) based on baseline score of the measure of interest, p=0•07; adjusted for baseline score, diagnosis, FEV 1 , SaO 2 %, p=0•072; control and intervention groups were imputed separately. In further post-hoc sensitivity analyses, fi rst excluding patients referred or identifi ed via palliative care services, and second excluding those with cancer (because of potential bias of palliative care eff ect), we noted 6 week mean mastery scores of, respectively, 4·18 (SD 1·2) in the intervention group and 3·54 (1·4) in the control group (p=0·043) and 4·19 (1·2) in the intervention group and 3·52 (1·3) in the control group (p=0·033).
We noted no signifi cant diff erences in patient-reported secondary outcomes between study groups at 6 weeks (table 2). For all items, except anxiety, the breathlessness support service group had better scores than the control group; this was largest, but not signifi cant, for the London Chest Activity of Daily Living questionnaire and breathlessness on exertion. Findings of pre-post analysis within groups (appendix p 8) showed signifi cant improvements in the breathlessness support service group between baseline and 6 weeks for seven outcomes: mastery, total quality of life score, dyspnoea, and emotion, assessed by Chronic Respiratory Disease Questionnaire, average breathlessness per 24 h, on exertion breathlessness per 24 h, and Palliative care Outcome Scale total score. No outcome showed deterioration. The control group had a signifi cant improvement between baseline and 6 weeks for only Palliative care Outcome Scale total score, and signifi cant deteriorations for London Chest Activity of Daily Living questionnaire and HADS.
We noted a signifi cant diff erence in survival for the whole sample that appeared early after randomisation (generalised Wilcoxon 3·90, p=0•048). Survival was similar between the study arms for patients with cancer, but signifi cantly diff erent for patients without cancer: all 42 patients without cancer in the breathlessness support service group were alive through to 6 months (180 days), of the 42 control patients without cancer at baseline, 38 were alive at 90 days, and 32 at 180 days (table 3 and fi gure 2). The standard care group received the breathlessness support service by 120 days.   At 6 weeks, we noted no signifi cant diff erences between total formal care costs in the two groups. 6 week mean costs were £1422 in the breathlessness support service group (bootstrapped 95% CI 897-2101) and £1408 in the control group (899-2023). Costs varied greatly between individuals.
Improved knowledge, confi dence, and insight into how to function despite breathlessness were identifi ed as potential mechanisms in the qualitative analysis through which the breathlessness support service improved patient mastery (appendix pp [10][11].

Discussion
This is the fi rst randomised trial of a breathlessness support service integrating palliative care and respiratory medicine, and the fi rst powered trial to test early integrated palliative care including patients without cancer (panel 2). The breathlessness support service integrated respiratory medicine, palliative care, physiotherapy, and occupational therapy for patients with advanced conditions and refractory breathlessness. The service responds to calls for earlier integration of palliative care including for patients without cancer. 10 At 6 weeks, the primary outcome, breathlessness mastery, improved more in the breathlessness support service group than in the standard care group. Qualitative data provided evidence of the breathlessness support service improving confi dence, function, and control over breathlessness. No secondary patient-reported outcomes were signifi cantly diff erent between groups, although there was evidence in the prepost analysis that the breathlessness support service group had improved activities of daily living and reduced breathlessness on exertion and depression. We recorded no harms of the breathlessness support service. The number of inpatient bed days and total formal care costs, on the basis of patient-reported total service use, were similar between groups.
Refractory breathlessness is a diffi cult clinical problem, usually the second most common symptom after pain in patients with advanced chronic disease, with high costs for society. 3 Oxygen has a role for individuals with severe hypoxaemia at rest or exercise desaturation, but is of little symptomatic value when patients are not hypoxic. 39 Lowdose, sustained-release opioids safely reduce breath lessness without respiratory depression, 40 but no other eff ective drugs exist. 1,3,41 Non-pharmacological treatments (eg, rollator devices, fan therapy, breathing control, and muscle strengthening) can provide benefi ts 16 as can multidisciplinary rehabilitation programmes, but in advanced disease, many patients are unable to attend or benefi t. 16,34 In this context, palliative care can have a role (panel 2), but robust trials are scarce.
Although the fi nding of improved mastery in patients in the breathlessness support service group might not be surprising, this service (integrated palliative and respiratory care) is not standard, and usual care did not achieve the same result. All patients had advanced and deteriorating disease, in the palliative phase of a progressive illness, in which breathlessness progressively increases up to death. Therefore, the fi nding of little change in our secondary outcomes is not surprising, especially ones such as spirometry. These data suggest that we included an appropriate group of patients.    Tick marks show censoring of data. Overall generalised Wilcoxon (Breslow) was 3·90, p=0·048. In subgroup analysis, this pattern was not recorded for patients with cancer (0·01, p=0·97; n=21); but it became more marked for patients with diseases other than cancer (6·04, p=0·01; n=84). The pattern was recorded for patients with chronic obstructive pulmonary disease and separately in those with interstitial lung disease, although numbers are small for those with interstitial lung disease. The American Thoracic Society defi nes breathlessness as "a subjective experience of breathing discomfort that consists of qualitatively distinct sensations that vary in intensity". 3 Like pain, breathlessness is a subjective experience resulting from complex interactions between pathological, physiological, and emotional elements. Therefore, it is diffi cult to choose the correct outcome to measure in breathlessness trials. We chose a well established quality of life measure's mastery domain, rather than levels of breathlessness, for several reasons. First, ratings of breathlessness levels are very variable and patients can have breathlessness attacks several times a day without constant breathlessness. Therefore, helping these patients to master their attacks might be more important than reducing the severity of one attack. Second, breathlessness is a limited endpoint because people perform activity to the highest level of breathlessness they can bear, but hope to do more before they reach that point. Breathlessness support services aimed to support and provide patients with coping strategies and interventions to help them master their breathlessness, while accepting that the disease cannot be cured and its natural history changed. Thus, the amount of perceived breathlessness mastery is probably a more important component of quality of life than is amount of breathlessness. As the qualitative results suggest, patients found this point important.
We found a diff erence in survival between study groups; patients in the control group had poorer survival in the early period of the study compared with patients in the breathlessness support service group. This diff erence was not found for patients with cancer, but was signifi cant for patients with diseases other than cancer, mostly in those with COPD and interstitial lung disease. We do not have reliable data for the longevity of the disease or prognosis before randomisation, which limits interpretation of this fi nding. However, our results support another trial of early palliative care, 31 although we are the fi rst to fi nd a survival diff erence for patients with diseases other than cancer. Therefore, these results need further exploration and testing in future trials, as does the optimum timing of the breathlessness support service.
The breathlessness support service had some similar components to the breathlessness intervention service developed in Cambridge, UK in the late 2000s, 15 including one point of entry, integration of palliative care with physiotherapy, some specifi c interventions, and education. However, there are diff erences between the services; we included assessment by respiratory medicine (a component valued by patients in the qualitative interviews), asked patients to attend outpatient clinics (the breathlessness intervention service is home based), and used the poem for crisis management.
Our study has limitations. We were only able to single mask the groups. Our primary outcome measure was subjective; patients who knew their study group could

Systematic review
In the past 4 years there have been calls for early integration of palliative care to support patients with advanced disease. 10,11 We searched the scientifi c literature for evidence by extending three systematic reviews and one narrative review that assessed early palliative care in hospitals and at home. [26][27][28][29][30] We updated their search terms and also searched for any studies with the keywords: palliative, integration, and early. Additionally, we searched for trials of service interventions for patients with breathlessness (or dyspnoea or dyspnea) in advanced diseases (chronic obstructive pulmonary disease [COPD] or interstitial lung disease or heart failure or cancer) and on refractory breathlessness. Although many qualitative and descriptive studies have been done, we identifi ed only four trials of early palliative care; three of these for patients with cancer. Temel and colleagues 31 and Bakitas and colleagues 32 found early palliative care signifi cantly improved quality of life; the recent Zimmermann and coworkers trial 28 found similar benefi ts. In non-cancer, there is one published phase 2 trial of early palliative care in patients with multiple sclerosis, which suggested early integrated palliative care improved symptom control. 33 Although many interventions and services have been proposed to help patients with chronic breathlessness, such as pulmonary rehabilitation, 34 and national initiatives such as IMPRESS aimed at the development of integrated population-based approaches to prevent, detect, and care for people aff ected by COPD, 35 there are few for patients with the more advanced stages of illness or refractory breathlessness. We identifi ed two randomised trials, one in progress. Bredin and colleagues 14 tested a nurse-led clinic for patients with lung cancer, not palliative care. A coauthor of this study, Booth, did observational work 15 and is currently trialling a community-based palliative care and physiotherapy service, the breathlessness intervention service, but this does not have out-patient clinics or include respiratory medicine. 36 In Canada, Rocker and Cook 37 developed the INSPIRED model of care for patients with advanced COPD to off er a more integrated approach to care, with personalised action plans, advance care planning, and palliative treatments. We could fi nd no randomised trials of INSPIRED but observational data suggest the model reduces repeat admissions and is welcomed by patients. Horton and colleagues 38 did an observational study to test the feasibility of home-based palliative care for patients and caregivers living with advanced COPD (30 patients enrolled and 13 provided outcome assessments).

Interpretation
In this randomised controlled trial of 105 patients with refractory breathlessness, we noted that patients who received an integrated palliative care and respiratory breathlessness support service had signifi cantly improved breathlessness mastery at 6 weeks. Mastery assessed patients' feeling of control over their breathlessness and its eff ects on quality of life and function, and was on average 16% higher for those patients receiving the breathlessness support service. The breathlessness support service did not show a signifi cant advantage for other secondary outcomes, although there was a tendency for improvement in the ability to undertake activities of daily living, lesser depression, and lower breathlessness on exertion. Our fi ndings were supported by qualitative data. We did not fi nd a diff erence in formal care costs. Like Temel and colleagues' trial, 31 survival was better for the group receiving early breathlessness support service than the control group, although in this instance for patients with diseases other than cancer. In both these studies, survival was a secondary outcome, which suggests further research is needed. Our study supports the early integration of palliative care with respiratory medicine in non-cancer (eg, COPD, interstitial lung disease, and heart failure), focused on a group with refractory breathlessness. By being based mainly in outpatient settings and for a short term, the breathlessness support service meets the Block and Billings criteria of being scalable. 11 As our trial was of one service, we suggest that the breathlessness support service warrants testing in multicentre randomised trials, and further studies comparing diff erent models and the timing of integration are needed.
have been subject to the placebo eff ect. However, participants were unaware that mastery was an endpoint because it was not emphasised in interviews and relevant questions were dispersed within the questionnaire. Additionally, the research nurse could have seen breathlessness support service equipment (eg, hand-held fan and information sheets) in the home, which could have biased their interviews. Our inclusion and exclusion criteria prevented extrapolation of study results to patients in the last month of life. Further, our outcome follow-up was short because of the fast-track nature of the trial. Although this short follow-up gave us acceptability from referrers and patients, it restricted our assessments, especially of care costs and long-term survival; the trial was not designed specifi cally to test for survival. We recruited from a small number of sites in urban areas where usual care at specialist centres was probably of an unusually good standard, with expert staff who were motivated to take part in this research. We were unable to contact more than a fi fth of patients screened and eligible for the study, and could not pursue this further because of data protection and ethics approval requirements. Therefore, we do not know how our recorded eff ects translate to other routine scenarios and settings. Some patients were identifi ed via palliative care services, which might have aff ected our results; however, the diff erence in our primary outcome remained when these eff ects were excluded.
Our primary outcome had an eff ect size of 0·44, smaller than that proposed in our sample size calculation. Puhan and colleagues 20 recommended an eff ect size of 0·7 for patient self-administered and 0·38 for intervieweradministered questionnaires. We fi nd it surprising that a diff erence in interviewer should make such a diff erence in eff ect size, although this might be related to less variation in interviewer-administered formats. However, perhaps we should have used the more conservative 0·38 in our sample size estimation.
This trial provides support for a more integrated approach to management of breathlessness within a breathlessness support service, which improves patient mastery without aff ecting overall care costs. The recorded improvement in survival needs further investigation. The breathlessness support service needs testing in multicentre, longer term trials including a wider range of urban and rural settings.